Seestar stack versus Siril stack

jgood

Having been led to believe that a manually stacked set of subs can always produce a better result than a Seestar S50 generated stack I've thrown myself fully into learning Siril and the in's and out's of the accompanying scripts. I've watched many videos and read tutorials on using Siril. What I'm finding is the Seestar stacked S50 subs are ALWAYS better than what I (an admitted novice) can produce via Siril. With Siril I am able to improve the image quality of nebulae and galaxies compared to the Seestar native output, but stars coming out of Siril look terrible. The Seestar on the other hand outputs stars that are remarkably good. They look real, sharp, properly color tinted with just the right amount of glow. I have been removing stars via StarNet in Siril, stretching the starless image and then recombining via Veralux StarComposer and my results are muddy, and artificial looking even though I'm not synthesizing anything.

So either I'm just crap at Siril—though I do have 25+ years working experience with imaging software and hardware so not a complete luddite—or ZWO software is a proprietary miracle App that turns so-so data into stellar (pun intended) images. Is anyone using Siril to stack and if so what are your approaches to improving stars?

3AC9bx99

I generally find the stacked fit file that Seestar produces is as good as the one produced running a script within Siril, and is the one I tend to use when processing in Siril. Before the Veralux scripts came out I always used to recombine stars into my starless image with StarNet using Modified Arcsinh Transformation stretch and Human Weighted Luminance as the colour stretch model. This helps prevent the stars from blowing out and preserves the colour.

jgood

Thanks for the info. I never actually considered using the Seestar .fits file. But that does make complete sense. I also never tried the Mod. Arcsinh Transformation on stars. That’s the problem with Youtube videos; they tend to divert you away from things the “influencers” aren’t promoting. I’ll give both suggestions a try today. Much appreciated.

3AC9bx99

There are other procedures within Siril to desaturate stars, fix elongated stars, reduce star size and remove aberrations. The YouTube channel by Deep Space Astro has a host of videos applicable for the Seestar within Siril. May be worth having a look. Example:-
https://www.youtube.com/watch?v=VNAcOcpubgQ

jgood

I’ve actually watched almost all of Deep Space Astro’s videos. He’s very good. At the moment I’m beginning to wonder if my real problem has more to do with posts on this site about the Seestar 3.0 App update introducing large amounts of noise into images. This is exactly what I’ve been seeing…massive amounts of grainy noise that affects everything. I’ve only had the 3.0, 3.0.1 and 3.0.2 App since I got my Seestar in late December. I’ll start with your initial suggestions and wait to see if there are updates to the App that provide fixes. Not having used the Seestar with prior software I was unaware the issue I’ve noted could be related. Thanks again!

3AC9bx99

jgood Yes it may well be any image problems may well be related to the latest App version. I have some problems too. Until this is sorted we can't move on. Lets hope ZWO come up with a fix soon. Clear skies!

francescoCh

@jgood Based on my experience with the S30, Siril is not good at denoising because the Seestar images are too noisy. Therefore my standard procedure always through goes Graxpert, with the very few exceptions being the big and shiny stuff (that would be Orion, Rosette, North America nebulas).
Graxpert does wonders in eliminating noise and revealing (...creating...) many hidden details especially on starless images. It's separate app but it's worth adding to the routine.
The other thing is that the default Seestar stacking script is not very good. Right now I only do the manual stacking with drizzle as the resampling algorithm. The standard script uses interpolation (debayer) instead which I don't particularly like.
You already know Veralux stuff and it's really good, especially for stretching. Denoising unfortunately is a bit disappointing.
Hope it helps

jgood

Thanks Francesco. These are all good suggestions. I have tried Graxpert and it did work well for both gradients and noise. I guess I'm still hunting for the combination of tools that gets me results that meet my expectations. Hearing what works for others is particularly helpful right now since I don't feel as though I have reproducible control. I'm working my way towards a reliable process that produces "good" results, and then from there I can begin to add tweaks to improve. And having no experience with any ZWO App software other than 3.0 and higher I keep wondering how much of the image quality is being affected by the clipping described elsewhere on the forum. I know my histograms look a lot like those being uploaded as examples of the problem. Time will tell. Thanks again for an excellent reply.

3AC9bx99

jgood After cropping the image I do a plate solve and then use Veralux background correction. Graxpert also works well. Do a SPCC colour calibration. I tend to use Cosmic Clarity Sharpen for both stars and object. Stretch with Veralux Hypermetric stretch. Apply Graxpert noise reduction at this point. The new Veralux Curves script works well after this. Can add HDR Multiscale and Veralux Revela to adjust contrast and finally colour saturation with Vectra. It is a case of trial and error between the one click Veralux scripts and the on board Siril programs. Veralux noise reduction is the only one I find a bit wanting, hence stick with Graxpert. Most of the work flow can be seen in video by Deep Space Astro.

jgood

Wow. You’ve both been exceptionally helpful. Thanks so much. I’ve got a lot to try now! And actually I do have one other Seestar related question. What is the orientation of the Seestar generated fits file in relation to the actual orientation of the stars in the sky? I’ve noted in some cases stacked fits files are horizontally inverted and other times they are not. Inversion appears to be correct when processed in Siril. Is that correct? But not when looking at stacked, raw fits files coming out of the Seestar?

3AC9bx99

The inversion is just a discrepancy between how the Seestar writes fits files and how Siril reads them. It does not affect the quality of the image. If the image appears inverted and plate solving doesn't flip it just flip them manually using the invert buttons along the bottom right in Siril before you start your processing.

jgood

So, if plate solving doesn’t automatically flip the image it means something else is wrong I’m guessing? I tried using two different S50 stacked fits files in Siril as starting points and neither of them would successfully plate solve. Are they already plate solved as they come out of the S50?

3AC9bx99

jgood Unlikely anything is wrong. The Seestar FITS header automatically puts in the RA and DEC parameters so unlikey to be that, but you can always enter the object identifier to be sure. Usually plate solving failures are down to insufficient stars or poor star quality. You can try to Downsample the image, tick the selection box and try again. Or you can select a region of interest, ROI, around a patch of dense stars, and try to plate solve again. I sometimes find I can still colour calibrate using SPCC even with a failed plate solve and continue my processing as normal. The stacked fits files are not plate solved straight out of the Seestar.

3AC9bx99

jgood I would also recommend you look at the YouTube video on plate solving in Siril by Deep Space Astro.

https://www.youtube.com/watch?v=iF_NaOIOH58&t=2s

Glabella

If this post is specifically about comparing your personal preference between Seestar app and your siril + others process then I would think you need to ensure you are aware of the differences in processing which might be difficult unless you have an in on the ZWO processing steps and parameters. My experience is my minimal hands on steps (meaning automation) with siril preprocess graxpert BE and NR then Siril post process with fixed setting for autostretch then half a minute of histogram transformation ( the only manual process I undertake anymore). I always find my method almost always produces more accurate images meaning easier to detect structures in my images than what Seestar app does. Now if I want to produce a prettier picture AND I don’t have lots of good data then I will usually use the SeeStar app photo after some further post post post processing LOL. If I have lots of data then I usually use my siril post processed png image after perhaps some minor post post processing. But that all depends if I think my audience wants to see pretty at the cost of closer to reality quality. Hey it’s a SeeStar S50 and unless you have lots off good to enter data then you won’t get much that people will go wow with. JMHO FWIW

jgood

Initially I was trying to understand if my novice attempts with Siril were the real problem, OR if ZWO had created some proprietary software to stack and post process S30/S50 subs that would be difficult to achieve via Siril. Since 3AC9bx99 provided some helpful suggestions I've seen a dramatic improvement in images I've processed using Siril as an adjunct. I've concluded that to a large degree, the problems I was having were due to my novice Siril skills. I'm improving but still not loving the stars after removing them via StarNet and replacing them with Veralux Starcomposer.

3AC9bx99

Not a bad image. Keep practising, try different approaches, there is no absolute right way. Lots of tutorials on line to help.